1 files changed, 207 insertions, 62 deletions

diff --git a/drivers/md/dm-bufio.c b/drivers/md/dm-bufio.c
index 2d519c223562..9c5ef818ca36 100644
--- a/drivers/md/dm-bufio.c
+++ b/drivers/md/dm-bufio.c
@@ -18,6 +18,7 @@
 #include <linux/module.h>
 #include <linux/rbtree.h>
 #include <linux/stacktrace.h>
+#include <linux/jump_label.h>
 
 #define DM_MSG_PREFIX "bufio"
 
@@ -81,6 +82,8 @@
  */
 struct dm_bufio_client {
 	struct mutex lock;
+	spinlock_t spinlock;
+	bool no_sleep;
 
 	struct list_head lru[LIST_SIZE];
 	unsigned long n_buffers[LIST_SIZE];
@@ -90,7 +93,6 @@ struct dm_bufio_client {
 	s8 sectors_per_block_bits;
 	void (*alloc_callback)(struct dm_buffer *);
 	void (*write_callback)(struct dm_buffer *);
-
 	struct kmem_cache *slab_buffer;
 	struct kmem_cache *slab_cache;
 	struct dm_io_client *dm_io;
@@ -108,7 +110,10 @@ struct dm_bufio_client {
 	int async_write_error;
 
 	struct list_head client_list;
+
 	struct shrinker shrinker;
+	struct work_struct shrink_work;
+	atomic_long_t need_shrink;
 };
 
 /*
@@ -158,23 +163,34 @@ struct dm_buffer {
 #endif
 };
 
+static DEFINE_STATIC_KEY_FALSE(no_sleep_enabled);
+
 /*----------------------------------------------------------------*/
 
 #define dm_bufio_in_request()	(!!current->bio_list)
 
 static void dm_bufio_lock(struct dm_bufio_client *c)
 {
-	mutex_lock_nested(&c->lock, dm_bufio_in_request());
+	if (static_branch_unlikely(&no_sleep_enabled) && c->no_sleep)
+		spin_lock_bh(&c->spinlock);
+	else
+		mutex_lock_nested(&c->lock, dm_bufio_in_request());
 }
 
 static int dm_bufio_trylock(struct dm_bufio_client *c)
 {
-	return mutex_trylock(&c->lock);
+	if (static_branch_unlikely(&no_sleep_enabled) && c->no_sleep)
+		return spin_trylock_bh(&c->spinlock);
+	else
+		return mutex_trylock(&c->lock);
 }
 
 static void dm_bufio_unlock(struct dm_bufio_client *c)
 {
-	mutex_unlock(&c->lock);
+	if (static_branch_unlikely(&no_sleep_enabled) && c->no_sleep)
+		spin_unlock_bh(&c->spinlock);
+	else
+		mutex_unlock(&c->lock);
 }
 
 /*----------------------------------------------------------------*/
@@ -256,12 +272,35 @@ static struct dm_buffer *__find(struct dm_bufio_client *c, sector_t block)
 		if (b->block == block)
 			return b;
 
-		n = (b->block < block) ? n->rb_left : n->rb_right;
+		n = block < b->block ? n->rb_left : n->rb_right;
 	}
 
 	return NULL;
 }
 
+static struct dm_buffer *__find_next(struct dm_bufio_client *c, sector_t block)
+{
+	struct rb_node *n = c->buffer_tree.rb_node;
+	struct dm_buffer *b;
+	struct dm_buffer *best = NULL;
+
+	while (n) {
+		b = container_of(n, struct dm_buffer, node);
+
+		if (b->block == block)
+			return b;
+
+		if (block <= b->block) {
+			n = n->rb_left;
+			best = b;
+		} else {
+			n = n->rb_right;
+		}
+	}
+
+	return best;
+}
+
 static void __insert(struct dm_bufio_client *c, struct dm_buffer *b)
 {
 	struct rb_node **new = &c->buffer_tree.rb_node, *parent = NULL;
@@ -276,8 +315,8 @@ static void __insert(struct dm_bufio_client *c, struct dm_buffer *b)
 		}
 
 		parent = *new;
-		new = (found->block < b->block) ?
-			&((*new)->rb_left) : &((*new)->rb_right);
+		new = b->block < found->block ?
+			&found->node.rb_left : &found->node.rb_right;
 	}
 
 	rb_link_node(&b->node, parent, new);
@@ -400,13 +439,13 @@ static void *alloc_buffer_data(struct dm_bufio_client *c, gfp_t gfp_mask,
 	 */
 	if (gfp_mask & __GFP_NORETRY) {
 		unsigned noio_flag = memalloc_noio_save();
-		void *ptr = __vmalloc(c->block_size, gfp_mask, PAGE_KERNEL);
+		void *ptr = __vmalloc(c->block_size, gfp_mask);
 
 		memalloc_noio_restore(noio_flag);
 		return ptr;
 	}
 
-	return __vmalloc(c->block_size, gfp_mask, PAGE_KERNEL);
+	return __vmalloc(c->block_size, gfp_mask);
 }
 
 /*
@@ -551,13 +590,12 @@ static void dmio_complete(unsigned long error, void *context)
 	b->end_io(b, unlikely(error != 0) ? BLK_STS_IOERR : 0);
 }
 
-static void use_dmio(struct dm_buffer *b, int rw, sector_t sector,
+static void use_dmio(struct dm_buffer *b, enum req_op op, sector_t sector,
 		     unsigned n_sectors, unsigned offset)
 {
 	int r;
 	struct dm_io_request io_req = {
-		.bi_op = rw,
-		.bi_op_flags = 0,
+		.bi_opf = op,
 		.notify.fn = dmio_complete,
 		.notify.context = b,
 		.client = b->c->dm_io,
@@ -585,11 +623,12 @@ static void bio_complete(struct bio *bio)
 {
 	struct dm_buffer *b = bio->bi_private;
 	blk_status_t status = bio->bi_status;
-	bio_put(bio);
+	bio_uninit(bio);
+	kfree(bio);
 	b->end_io(b, status);
 }
 
-static void use_bio(struct dm_buffer *b, int rw, sector_t sector,
+static void use_bio(struct dm_buffer *b, enum req_op op, sector_t sector,
 		    unsigned n_sectors, unsigned offset)
 {
 	struct bio *bio;
@@ -600,16 +639,14 @@ static void use_bio(struct dm_buffer *b, int rw, sector_t sector,
 	if (unlikely(b->c->sectors_per_block_bits < PAGE_SHIFT - SECTOR_SHIFT))
 		vec_size += 2;
 
-	bio = bio_kmalloc(GFP_NOWAIT | __GFP_NORETRY | __GFP_NOWARN, vec_size);
+	bio = bio_kmalloc(vec_size, GFP_NOWAIT | __GFP_NORETRY | __GFP_NOWARN);
 	if (!bio) {
 dmio:
-		use_dmio(b, rw, sector, n_sectors, offset);
+		use_dmio(b, op, sector, n_sectors, offset);
 		return;
 	}
-
+	bio_init(bio, b->c->bdev, bio->bi_inline_vecs, vec_size, op);
 	bio->bi_iter.bi_sector = sector;
-	bio_set_dev(bio, b->c->bdev);
-	bio_set_op_attrs(bio, rw, 0);
 	bio->bi_end_io = bio_complete;
 	bio->bi_private = b;
 
@@ -631,7 +668,21 @@ dmio:
 	submit_bio(bio);
 }
 
-static void submit_io(struct dm_buffer *b, int rw, void (*end_io)(struct dm_buffer *, blk_status_t))
+static inline sector_t block_to_sector(struct dm_bufio_client *c, sector_t block)
+{
+	sector_t sector;
+
+	if (likely(c->sectors_per_block_bits >= 0))
+		sector = block << c->sectors_per_block_bits;
+	else
+		sector = block * (c->block_size >> SECTOR_SHIFT);
+	sector += c->start;
+
+	return sector;
+}
+
+static void submit_io(struct dm_buffer *b, enum req_op op,
+		      void (*end_io)(struct dm_buffer *, blk_status_t))
 {
 	unsigned n_sectors;
 	sector_t sector;
@@ -639,13 +690,9 @@ static void submit_io(struct dm_buffer *b, int rw, void (*end_io)(struct dm_buff
 
 	b->end_io = end_io;
 
-	if (likely(b->c->sectors_per_block_bits >= 0))
-		sector = b->block << b->c->sectors_per_block_bits;
-	else
-		sector = b->block * (b->c->block_size >> SECTOR_SHIFT);
-	sector += b->c->start;
+	sector = block_to_sector(b->c, b->block);
 
-	if (rw != REQ_OP_WRITE) {
+	if (op != REQ_OP_WRITE) {
 		n_sectors = b->c->block_size >> SECTOR_SHIFT;
 		offset = 0;
 	} else {
@@ -664,9 +711,9 @@ static void submit_io(struct dm_buffer *b, int rw, void (*end_io)(struct dm_buff
 	}
 
 	if (b->data_mode != DATA_MODE_VMALLOC)
-		use_bio(b, rw, sector, n_sectors, offset);
+		use_bio(b, op, sector, n_sectors, offset);
 	else
-		use_dmio(b, rw, sector, n_sectors, offset);
+		use_dmio(b, op, sector, n_sectors, offset);
 }
 
 /*----------------------------------------------------------------
@@ -748,7 +795,8 @@ static void __make_buffer_clean(struct dm_buffer *b)
 {
 	BUG_ON(b->hold_count);
 
-	if (!b->state)	/* fast case */
+	/* smp_load_acquire() pairs with read_endio()'s smp_mb__before_atomic() */
+	if (!smp_load_acquire(&b->state))	/* fast case */
 		return;
 
 	wait_on_bit_io(&b->state, B_READING, TASK_UNINTERRUPTIBLE);
@@ -768,6 +816,10 @@ static struct dm_buffer *__get_unclaimed_buffer(struct dm_bufio_client *c)
 		BUG_ON(test_bit(B_WRITING, &b->state));
 		BUG_ON(test_bit(B_DIRTY, &b->state));
 
+		if (static_branch_unlikely(&no_sleep_enabled) && c->no_sleep &&
+		    unlikely(test_bit_acquire(B_READING, &b->state)))
+			continue;
+
 		if (!b->hold_count) {
 			__make_buffer_clean(b);
 			__unlink_buffer(b);
@@ -776,6 +828,9 @@ static struct dm_buffer *__get_unclaimed_buffer(struct dm_bufio_client *c)
 		cond_resched();
 	}
 
+	if (static_branch_unlikely(&no_sleep_enabled) && c->no_sleep)
+		return NULL;
+
 	list_for_each_entry_reverse(b, &c->lru[LIST_DIRTY], lru_list) {
 		BUG_ON(test_bit(B_READING, &b->state));
 
@@ -1004,7 +1059,7 @@ found_buffer:
 	 * If the user called both dm_bufio_prefetch and dm_bufio_get on
 	 * the same buffer, it would deadlock if we waited.
 	 */
-	if (nf == NF_GET && unlikely(test_bit(B_READING, &b->state)))
+	if (nf == NF_GET && unlikely(test_bit_acquire(B_READING, &b->state)))
 		return NULL;
 
 	b->hold_count++;
@@ -1164,7 +1219,7 @@ void dm_bufio_release(struct dm_buffer *b)
 		 * invalid buffer.
 		 */
 		if ((b->read_error || b->write_error) &&
-		    !test_bit(B_READING, &b->state) &&
+		    !test_bit_acquire(B_READING, &b->state) &&
 		    !test_bit(B_WRITING, &b->state) &&
 		    !test_bit(B_DIRTY, &b->state)) {
 			__unlink_buffer(b);
@@ -1307,8 +1362,7 @@ EXPORT_SYMBOL_GPL(dm_bufio_write_dirty_buffers);
 int dm_bufio_issue_flush(struct dm_bufio_client *c)
 {
 	struct dm_io_request io_req = {
-		.bi_op = REQ_OP_WRITE,
-		.bi_op_flags = REQ_PREFLUSH | REQ_SYNC,
+		.bi_opf = REQ_OP_WRITE | REQ_PREFLUSH | REQ_SYNC,
 		.mem.type = DM_IO_KMEM,
 		.mem.ptr.addr = NULL,
 		.client = c->dm_io,
@@ -1326,6 +1380,29 @@ int dm_bufio_issue_flush(struct dm_bufio_client *c)
 EXPORT_SYMBOL_GPL(dm_bufio_issue_flush);
 
 /*
+ * Use dm-io to send a discard request to flush the device.
+ */
+int dm_bufio_issue_discard(struct dm_bufio_client *c, sector_t block, sector_t count)
+{
+	struct dm_io_request io_req = {
+		.bi_opf = REQ_OP_DISCARD | REQ_SYNC,
+		.mem.type = DM_IO_KMEM,
+		.mem.ptr.addr = NULL,
+		.client = c->dm_io,
+	};
+	struct dm_io_region io_reg = {
+		.bdev = c->bdev,
+		.sector = block_to_sector(c, block),
+		.count = block_to_sector(c, count),
+	};
+
+	BUG_ON(dm_bufio_in_request());
+
+	return dm_io(&io_req, 1, &io_reg, NULL);
+}
+EXPORT_SYMBOL_GPL(dm_bufio_issue_discard);
+
+/*
  * We first delete any other buffer that may be at that new location.
  *
  * Then, we write the buffer to the original location if it was dirty.
@@ -1401,6 +1478,14 @@ retry:
 }
 EXPORT_SYMBOL_GPL(dm_bufio_release_move);
 
+static void forget_buffer_locked(struct dm_buffer *b)
+{
+	if (likely(!b->hold_count) && likely(!smp_load_acquire(&b->state))) {
+		__unlink_buffer(b);
+		__free_buffer_wake(b);
+	}
+}
+
 /*
  * Free the given buffer.
  *
@@ -1414,15 +1499,36 @@ void dm_bufio_forget(struct dm_bufio_client *c, sector_t block)
 	dm_bufio_lock(c);
 
 	b = __find(c, block);
-	if (b && likely(!b->hold_count) && likely(!b->state)) {
-		__unlink_buffer(b);
-		__free_buffer_wake(b);
-	}
+	if (b)
+		forget_buffer_locked(b);
 
 	dm_bufio_unlock(c);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_forget);
 
+void dm_bufio_forget_buffers(struct dm_bufio_client *c, sector_t block, sector_t n_blocks)
+{
+	struct dm_buffer *b;
+	sector_t end_block = block + n_blocks;
+
+	while (block < end_block) {
+		dm_bufio_lock(c);
+
+		b = __find_next(c, block);
+		if (b) {
+			block = b->block + 1;
+			forget_buffer_locked(b);
+		}
+
+		dm_bufio_unlock(c);
+
+		if (!b)
+			break;
+	}
+
+}
+EXPORT_SYMBOL_GPL(dm_bufio_forget_buffers);
+
 void dm_bufio_set_minimum_buffers(struct dm_bufio_client *c, unsigned n)
 {
 	c->minimum_buffers = n;
@@ -1437,7 +1543,11 @@ EXPORT_SYMBOL_GPL(dm_bufio_get_block_size);
 
 sector_t dm_bufio_get_device_size(struct dm_bufio_client *c)
 {
-	sector_t s = i_size_read(c->bdev->bd_inode) >> SECTOR_SHIFT;
+	sector_t s = bdev_nr_sectors(c->bdev);
+	if (s >= c->start)
+		s -= c->start;
+	else
+		s = 0;
 	if (likely(c->sectors_per_block_bits >= 0))
 		s >>= c->sectors_per_block_bits;
 	else
@@ -1446,6 +1556,12 @@ sector_t dm_bufio_get_device_size(struct dm_bufio_client *c)
 }
 EXPORT_SYMBOL_GPL(dm_bufio_get_device_size);
 
+struct dm_io_client *dm_bufio_get_dm_io_client(struct dm_bufio_client *c)
+{
+	return c->dm_io;
+}
+EXPORT_SYMBOL_GPL(dm_bufio_get_dm_io_client);
+
 sector_t dm_bufio_get_block_number(struct dm_buffer *b)
 {
 	return b->block;
@@ -1522,8 +1638,9 @@ static void drop_buffers(struct dm_bufio_client *c)
  */
 static bool __try_evict_buffer(struct dm_buffer *b, gfp_t gfp)
 {
-	if (!(gfp & __GFP_FS)) {
-		if (test_bit(B_READING, &b->state) ||
+	if (!(gfp & __GFP_FS) ||
+	    (static_branch_unlikely(&no_sleep_enabled) && b->c->no_sleep)) {
+		if (test_bit_acquire(B_READING, &b->state) ||
 		    test_bit(B_WRITING, &b->state) ||
 		    test_bit(B_DIRTY, &b->state))
 			return false;
@@ -1549,8 +1666,7 @@ static unsigned long get_retain_buffers(struct dm_bufio_client *c)
 	return retain_bytes;
 }
 
-static unsigned long __scan(struct dm_bufio_client *c, unsigned long nr_to_scan,
-			    gfp_t gfp_mask)
+static void __scan(struct dm_bufio_client *c)
 {
 	int l;
 	struct dm_buffer *b, *tmp;
@@ -1561,42 +1677,58 @@ static unsigned long __scan(struct dm_bufio_client *c, unsigned long nr_to_scan,
 
 	for (l = 0; l < LIST_SIZE; l++) {
 		list_for_each_entry_safe_reverse(b, tmp, &c->lru[l], lru_list) {
-			if (__try_evict_buffer(b, gfp_mask))
+			if (count - freed <= retain_target)
+				atomic_long_set(&c->need_shrink, 0);
+			if (!atomic_long_read(&c->need_shrink))
+				return;
+			if (__try_evict_buffer(b, GFP_KERNEL)) {
+				atomic_long_dec(&c->need_shrink);
 				freed++;
-			if (!--nr_to_scan || ((count - freed) <= retain_target))
-				return freed;
+			}
 			cond_resched();
 		}
 	}
-	return freed;
 }
 
-static unsigned long
-dm_bufio_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
+static void shrink_work(struct work_struct *w)
+{
+	struct dm_bufio_client *c = container_of(w, struct dm_bufio_client, shrink_work);
+
+	dm_bufio_lock(c);
+	__scan(c);
+	dm_bufio_unlock(c);
+}
+
+static unsigned long dm_bufio_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
 {
 	struct dm_bufio_client *c;
-	unsigned long freed;
 
 	c = container_of(shrink, struct dm_bufio_client, shrinker);
-	if (sc->gfp_mask & __GFP_FS)
-		dm_bufio_lock(c);
-	else if (!dm_bufio_trylock(c))
-		return SHRINK_STOP;
+	atomic_long_add(sc->nr_to_scan, &c->need_shrink);
+	queue_work(dm_bufio_wq, &c->shrink_work);
 
-	freed  = __scan(c, sc->nr_to_scan, sc->gfp_mask);
-	dm_bufio_unlock(c);
-	return freed;
+	return sc->nr_to_scan;
 }
 
-static unsigned long
-dm_bufio_shrink_count(struct shrinker *shrink, struct shrink_control *sc)
+static unsigned long dm_bufio_shrink_count(struct shrinker *shrink, struct shrink_control *sc)
 {
 	struct dm_bufio_client *c = container_of(shrink, struct dm_bufio_client, shrinker);
 	unsigned long count = READ_ONCE(c->n_buffers[LIST_CLEAN]) +
 			      READ_ONCE(c->n_buffers[LIST_DIRTY]);
 	unsigned long retain_target = get_retain_buffers(c);
+	unsigned long queued_for_cleanup = atomic_long_read(&c->need_shrink);
+
+	if (unlikely(count < retain_target))
+		count = 0;
+	else
+		count -= retain_target;
+
+	if (unlikely(count < queued_for_cleanup))
+		count = 0;
+	else
+		count -= queued_for_cleanup;
 
-	return (count < retain_target) ? 0 : (count - retain_target);
+	return count;
 }
 
 /*
@@ -1605,7 +1737,8 @@ dm_bufio_shrink_count(struct shrinker *shrink, struct shrink_control *sc)
 struct dm_bufio_client *dm_bufio_client_create(struct block_device *bdev, unsigned block_size,
 					       unsigned reserved_buffers, unsigned aux_size,
 					       void (*alloc_callback)(struct dm_buffer *),
-					       void (*write_callback)(struct dm_buffer *))
+					       void (*write_callback)(struct dm_buffer *),
+					       unsigned int flags)
 {
 	int r;
 	struct dm_bufio_client *c;
@@ -1635,12 +1768,18 @@ struct dm_bufio_client *dm_bufio_client_create(struct block_device *bdev, unsign
 	c->alloc_callback = alloc_callback;
 	c->write_callback = write_callback;
 
+	if (flags & DM_BUFIO_CLIENT_NO_SLEEP) {
+		c->no_sleep = true;
+		static_branch_inc(&no_sleep_enabled);
+	}
+
 	for (i = 0; i < LIST_SIZE; i++) {
 		INIT_LIST_HEAD(&c->lru[i]);
 		c->n_buffers[i] = 0;
 	}
 
 	mutex_init(&c->lock);
+	spin_lock_init(&c->spinlock);
 	INIT_LIST_HEAD(&c->reserved_buffers);
 	c->need_reserved_buffers = reserved_buffers;
 
@@ -1687,11 +1826,15 @@ struct dm_bufio_client *dm_bufio_client_create(struct block_device *bdev, unsign
 		__free_buffer_wake(b);
 	}
 
+	INIT_WORK(&c->shrink_work, shrink_work);
+	atomic_long_set(&c->need_shrink, 0);
+
 	c->shrinker.count_objects = dm_bufio_shrink_count;
 	c->shrinker.scan_objects = dm_bufio_shrink_scan;
 	c->shrinker.seeks = 1;
 	c->shrinker.batch = 0;
-	r = register_shrinker(&c->shrinker);
+	r = register_shrinker(&c->shrinker, "md-%s:(%u:%u)", slab_name,
+			      MAJOR(bdev->bd_dev), MINOR(bdev->bd_dev));
 	if (r)
 		goto bad;
 
@@ -1732,6 +1875,7 @@ void dm_bufio_client_destroy(struct dm_bufio_client *c)
 	drop_buffers(c);
 
 	unregister_shrinker(&c->shrinker);
+	flush_work(&c->shrink_work);
 
 	mutex_lock(&dm_bufio_clients_lock);
 
@@ -1762,6 +1906,8 @@ void dm_bufio_client_destroy(struct dm_bufio_client *c)
 	kmem_cache_destroy(c->slab_buffer);
 	dm_io_client_destroy(c->dm_io);
 	mutex_destroy(&c->lock);
+	if (c->no_sleep)
+		static_branch_dec(&no_sleep_enabled);
 	kfree(c);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_client_destroy);
@@ -1965,7 +2111,6 @@ static void __exit dm_bufio_exit(void)
 	int bug = 0;
 
 	cancel_delayed_work_sync(&dm_bufio_cleanup_old_work);
-	flush_workqueue(dm_bufio_wq);
 	destroy_workqueue(dm_bufio_wq);
 
 	if (dm_bufio_client_count) {